A computer-implemented method according to one embodiment includes receiving voice input, parsing the voice input for determining words in the voice input, recognizing at least some of the words, identifying an accent of a speaker of the voice input, evaluating a word accent of at least some of the recognized words based on the identified accent of the speaker and analysis of the recognized words, generating an evaluated accent score for each of the at least some of the recognized words based on the evaluated word accent, identifying at least one of the recognized words as an accent word based on the evaluated accent score thereof, and outputting a recommended correct pronunciation of the accent word or words.